Uniform representations of structured patterns and an algorithm for the induction of contingency-response rules

摘要: Many events (patterns) may be described by structural (conjunctive relational) representations, and general computational behavior represented in terms of a set grammatical rules (productions, transformations) relating two such event representations as contingency response components. Uniform graphs descriptions are introduced. An abstraction uniform corresponds to common subgraph the corresponding graphs. Every rule F = [(∀x1 ,…, xn) C(x1 ⇒ R(x1 xn)] which can induced from training I {(Ci , Ri): i= 1,…, N} contingency—response (input—output) pairs is identified with causal inferences Ci Ri. A learning problem formulated for three cases distinguishable on basis if how substitutions input output patterns made. Category (unary) n-ary predicate this framework discussed. Examples applications drawn domains transformational grammar. The properties (both desirable undesirable) proposed approach differences between it previous approaches also considered.